🚢 Maritime Incident Report Generator (RAG-based)

This project is an end-to-end Retrieval-Augmented Generation (RAG) system that generates structured maritime incident reports and plans of action based on historical incident data and Standard Operating Procedure (SOP) documents.

It leverages:

• 🔍 FAISS + Sentence Transformers for semantic retrieval of relevant SOPs
• 🧠 OpenAI LLM (via LangChain) for intelligent report generation
• ⚙️ FastAPI backend to handle RAG processing and embedding pipeline
• 💡 Streamlit frontend for a simple UI with one-click summarization

---

🧩 Features

• Accepts multiple maritime incident entries (JSON format)
• Retrieves the most relevant SOP content using semantic similarity
• Summarizes incidents and SOPs into a unified report
• Generates a clear Incident Report and Plan of Action
• Provides a Streamlit UI for easy interaction

---

🗂️ Project Structure

rag_incident_report/
├── app/
│   ├── main.py                # FastAPI app entry
│   ├── api.py                 # API routes
│   └── services/
│       ├── rag.py             # RAG logic (embedding, retrieval, generation)
│       └── sop_loader.py      # SOP loading from PDFs/DOCX/TXT
│   
│               
├── frontend/
│   └── streamlit_app.py       # Streamlit frontend app
├── data/
│   └── sops/                  # Folder containing SOP documents
├── .env                       # API keys and environment configs
└── schema.py                  # Pydantic request models                     
├── requirements.txt
└── README.md

---

🚀 Getting Started

1. Clone the repository

git clone https://github.com/yash-rewalia/RAG_Marine_Industry_safety.git
cd 'your-folder-name'

2. Install dependencies

pip install -r requirements.txt

3. Add your OpenAI API key

Create a .env file:

OPENAI_API_KEY=your_openai_api_key_here

4. Place your SOPs

Put .pdf, .docx, or .txt files inside the data/sops/ folder.

---

🧪 Running the App

Start the FastAPI backend:

uvicorn app.main:app --reload

Run the Streamlit frontend:

streamlit run frontend/streamlit_app.py

---

📸 UI Preview

Streamlit shows:

• Incident Summary
• Retrieved SOP Context
• Generated Report & Plan of Action

---

📄 Sample Input Format

{
  "incidents": [
    {
      "event": "Fire in engine room",
      "eventdate": "12-07-2023",
      "longdesc": "A fire broke out in the engine room due to overheating of machinery.",
      "veseltype": "Tanker",
      "rootcause": "Lack of maintenance"
    }
  ]
}

---

🙌 Acknowledgements

• OpenAI
• LangChain
• Sentence-Transformers
• FAISS
• Streamlit
• FastAPI

---

Made with ❤️ for safer seas and smarter incident handling.